European IT departments are undergoing one of the most significant structural transformations in a generation. Agentic AI, the breed of artificial intelligence that does not merely assist but actively reasons and executes, is turning traditional IT functions into something that looks far more like human resources. Teams that once managed servers, licences, and service desks are now tasked with hiring, onboarding, and performance-managing a parallel digital workforce that neither sleeps nor requires a pension.
The Rise of IT as Agent Manager
At Mobile World Congress in Barcelona earlier this year, industry leaders drew a pointed parallel between the current agentic AI shift and the microservices revolution that broke apart monolithic enterprise applications a decade ago. HPE's Bryan Thompson, VP for GreenLake product management, made the architectural logic plain: "Agentic AI is the next step in breaking apart and solving problems. There are opportunities to leverage these types of models and break them into almost like a microservice type of approach to tackle them, breaking them apart into specialised services."
The comparison is instructive for European IT leaders because the microservices transition, for all its turbulence, ultimately delivered more resilient, scalable infrastructure. Those who moved early captured the efficiency gains; those who waited spent years playing catch-up. The same dynamic is now in play with agentic AI, and the stakes are considerably higher.
Matthias Holweg, Professor of Operations Management at the University of Oxford's Said Business School and a prominent voice on AI-driven organisational change, has argued publicly that the shift toward autonomous agents demands new governance structures, not just new tooling. His position is blunt: enterprises that bolt agentic capabilities onto existing IT governance frameworks will encounter serious accountability gaps, particularly in regulated sectors such as financial services and healthcare.
Independent Execution Changes Everything
What distinguishes agentic AI from every enterprise software category that preceded it is its capacity for independent thought and action. Deloitte's Abdi Goodarzi, Head of Gen AI Products, Innovations and New Businesses, has been direct about why this matters: "Until now, we've never had a technology that could ideate, or execute independently. Just think about that statement, and any other software package solution you've ever dealt with. None of them could independently execute any of it."
That single observation carries enormous weight for IT departments in Germany, France, the Netherlands, and across the UK. Managing a system that can decide and act is categorically different from managing one that merely responds to instructions. The traditional IT skill set, centred on configuration, monitoring, and patching, is necessary but no longer sufficient.
Nvidia's Fred Devoir, Global Head of Solution Architecture for Telco, frames the workflow implications in practical terms: "We take componentry and put it together into a RESTful architecture. Nvidia was able to optimise those with our microservices, and then bring together those microservices into blueprints to give a very quick time to value or time to first results." The modular, composable design philosophy he describes maps directly onto how a modern HR function assembles specialist contractors into project teams, briefing them, monitoring their output, and re-deploying them as priorities shift.
Goodarzi raises a subtler challenge that European organisations would do well to take seriously: "Agents don't have emotions. How do you incorporate the emotions that will be part of the execution of the work?" In customer-facing deployments, in healthcare triage, in legal document review, the absence of emotional calibration is not a minor technical detail. It is a design and governance problem that sits squarely in the lap of whoever manages the agent workforce.
Solving the Data Silo Problem Without Moving the Data
One of the most persistent obstacles to enterprise AI adoption in Europe is the data silo. Legacy ERP systems, disparate CRM platforms, and siloed operational databases make centralising data for AI training both expensive and, under the General Data Protection Regulation, legally complex. Agentic AI offers a compelling architectural answer: take the intelligence to the data, rather than the reverse.
Devoir explains the model: "Instead of having to bring all your data to the AI, you're taking the AI to the data. When you make a service call, it actually asks all those data agents for a response and collates that data into a model." For European enterprises operating across multiple jurisdictions with different data residency requirements, this distributed approach is not merely convenient; it may be the only compliant option.
The practical benefits of this distributed architecture include the following:
- Eliminates costly data migration projects by working with existing systems in place
- Maintains data sovereignty and compliance requirements across EU member state jurisdictions
- Reduces latency by processing information closer to its source
- Allows incremental deployment without disrupting core business operations
- Enables specialised agents for different data types and business functions
The European AI Act, which began its phased enforcement in 2024, reinforces the importance of this approach. High-risk AI systems, including those used in employment, credit, and critical infrastructure, face stringent transparency and auditability requirements. A distributed, agent-based architecture that processes data in place rather than aggregating it into a central model is considerably easier to audit and to explain to a regulator.
Trust, Verification, and the Probabilistic Gap
Traditional enterprise IT is built on determinism: given the same inputs, the system produces the same output, every time. Agentic AI operates on fundamentally different principles. Its outputs are probabilistic, shaped by training, context, and the particular framing of each query. For IT departments accustomed to binary pass or fail diagnostics, this represents a genuine conceptual shift.
Goodarzi is candid about the verification challenge: "Am I dealing with the right data? Am I dealing with the right results? Agentic AI is designed around probabilistic technologies. So you get the best probable answer because you have trained agents with a lot of knowledge on how to digest the data and make a decision."
The European approach to this challenge is taking shape through the work of the AI Office, established under the European Commission in early 2024 to oversee implementation of the AI Act. The Office has been clear that organisations deploying high-risk agentic systems must maintain meaningful human oversight and cannot simply point to probabilistic accuracy rates as a substitute for explainability. For IT leaders, this means investing in new quality assurance frameworks, audit trails, and performance monitoring tools that go well beyond the uptime dashboards of traditional operations.
The management differences between legacy IT and agent management are stark. Where traditional systems are deployed through installation and configuration, agents require training, fine-tuning, and onboarding. Performance monitoring shifts from system metrics and uptime to decision quality and learning progress. Troubleshooting moves from error logs and debugging to behaviour analysis and bias detection. Compliance evolves from access controls and audit logs to explainability requirements and ethical guardrails.
New Roles, New Competencies, New Organisational Structures
The HR analogy that runs through this transformation is more than rhetorical convenience. It has direct implications for how European enterprises structure their IT functions and what they recruit for. The competencies required to manage an agent workforce overlap substantially with those of a progressive HR department: understanding motivation and incentive design (in the agent context, training objectives and reward functions), managing performance reviews (continuous monitoring and retraining cycles), handling misconduct (bias detection and behavioural correction), and ensuring legal compliance (the AI Act's conformity assessments and incident reporting obligations).
Holweg's research suggests that organisations which invest early in hybrid roles combining technical AI expertise with organisational design capabilities will outperform those that treat agent deployment as a purely technical project. The winners will not simply be those with the most sophisticated agents; they will be those with the clearest governance structures and the most coherent management philosophies for supervising autonomous systems at scale.
Dedicated "Agent Operations" functions, sitting at the intersection of IT, HR, legal, and compliance, are already emerging at a handful of large European enterprises. By 2026, such roles are likely to be a standard feature of any organisation running more than a handful of production AI agents. The question for European IT leaders is not whether to build this capability, but how quickly.
The Human Element Remains Non-Negotiable
None of this means human employees are about to be displaced wholesale. The more accurate picture is a hybrid workforce in which humans and agents collaborate on complementary tasks. Agents handle routine, data-intensive, and analytical work at speed and scale. Humans provide the emotional intelligence, contextual judgement, creative problem-solving, and ethical oversight that agents structurally cannot replicate.
For European enterprises, that division of labour also reflects regulatory reality. The AI Act's human oversight requirements for high-risk applications are not optional add-ons; they are legal obligations. Building a management culture that treats human oversight as integral rather than burdensome will be a competitive differentiator, not just a compliance exercise.
The transformation of IT into the HR function for AI agents is not a future scenario. It is happening now, in enterprise IT departments from Amsterdam to Milan, from Munich to Manchester. The organisations that recognise this shift early, and invest accordingly in new skills, new governance frameworks, and new management philosophies, will set the pace for everyone else.
Comments
Sign in to join the conversation. Be civil, be specific, link your sources.